Classification of Internet Traffic Data Using Ensemble Method

Accurate traffic classification is critical in network security and traffic engineering. Traditional methods based on port numbers and payload have proved to be ineffective in terms of dynamic port allocation and packet encapsulation. These methods also fail if the data is encrypted. In this work, we propose a ensemble method using (a) extra-tree-based feature selection, (b) support vector machines (SVMs) for classification of Internet traffic using various kernels, and finally (c) ensemble of SVM classifier using major voting technique. The method classifies the Internet traffic into broad application categories according to the network flow parameters obtained from the packet headers. We first compare three types of SVM kernels, i.e., linear, polynomial, and radial basis function (RBF) kernels. Later, we combine all the three kernels through majority voting (ensemble) method. In most of the cases, ensemble method gives better result compared with all other kernel methods.